Practical automation and interpretation of quantitative assays of antibodies to therapeutic proteins, illustrated with human growth hormone.

Abstract

To quantify concentrations of anti-growth hormone antibody in less than 600 serum samples by radioimmunoassay, we devised a system ("BOLAC") to process data and to execute the curve-fitting program LIGAND-PC automatically with an IBM PC-compatible computer. We fit data from each sample to four binding and one or two-antibody binding sites. Total antibody concentration is then calculated from the model that is statistically "best". This process occasionally selects a two-binding-site model that severely overestimates the antibody concentration. Errors of this kind are discarded by constraining the product of the second-site antibody's affinity and its concentration to exceed a minimum value (0.05). We evaluated the performance of the BOLAC system by assaying controls and by using computer simulations to demonstrate the high confidence levels attainable in estimation of antibody concentrations. Between-assay variability (CV) was less than 25%, and analytical recovery exceeded 90%. These figures are acceptable for an assay based on curve-fitting of competitive radioimmunoassay data, allowing clinically relevant assessments of antibody responses in patient's samples. The advantages of the BOLAC system include high throughput and the reporting of results in absolute units of affinities and concentrations.